Pidsdps0805 MF Rural Households

Philippine Institute for Development Studies
Surian sa mga Pag-aaral Pangkaunlaran ng Pilipinas
Impact of Microfinance
on Rural Households in the Philippines
Toshio Kondo, Aniceto Orbeta Jr.
Clarence Dingcong, and Christine Infantado
DISCUSSION PAPER SERIES NO. 2008-05
The PIDS Discussion Paper Series

constitutes studies that are preliminary and
subject to further revisions. They are being circulated in a limited number of copies only for purposes of soliciting comments and suggestions for further refinements. The studies under the Series are
unedited and unreviewed.
The views and opinions expressed
are those of the author(s) and do not necessarily reflect those of the Institute.
Not for quotation without permission
from the author(s) and the Institute.
February 2008
For comments, suggestions or further inquiries please contact:
The Research Information Staff, Philippine Institute for Development Studies
5th Floor, NEDA sa Makati Building, 106 Amorsolo Street, Legaspi Village, Makati City, Philippines
Tel Nos: (63-2) 8942584 and 8935705; Fax No: (63-2) 8939589; E-mail: publications@pids.gov.ph
Or visit our website at http://www.pids.gov.ph
IMPACT OF MICROFINANCE ON RURAL HOUSEHOLDS

IN THE PHILIPPINES
Toshio Kondo, Aniceto Orbeta, Jr., Clarence Dingcong, and Christine Infantado
Abstract
This paper reports on the impact evaluation study of the Rural Microenterprise Finance
Project (RMFP) in the Philippines. RMFP aimed to support efforts of the Government of
the Philippines to strengthen rural financial institutions by assisting organizations that
employed the Grameen Bank Approach (GBA) in providing credit to the poor. The
project was implemented by the Peoples Credit and Finance Corporation (PCFC) and
funded by the Asian Development Bank.
The evaluation uses a quasi-experimental design with incoming clients of randomly
selected participating microfinance institutions as the comparison group. An important
innovation in the study is the inclusion of the appropriate number of former clients
among the treatment group. Qualified non-participating households provide the control
for area effects. The impact estimation uses the difference-in-difference estimation
technique which effectively controls for the known sources of biases namely: nonrandom program participation (sample selection), non-random program placement, and
non-random drop-out. The survey enumerated some 2,200 households divided evenly
between treatment and comparison areas. It covered 116 villages spread throughout the
three groups of islands (Luzon, Visayas, and Mindanao) and 38 microfinance institutions
consisting of three types - banks, cooperatives, and non-government organizations.
The survey results show the program appears to be hitting only a limited number of the
intended target as majority of the existing clients and the incoming clients are found to
be not poor according to official definition. The estimation results show a mildly
significant positive impact on per capita income, per capita total expenditure and per
capita food expenditure of loan availability. This impact, however, was found to be
regressive negative on poorer households and positive only for households in the
richest quartile. The program has enabled participants to reduce dependence on
presumably higher-priced non-program loans as well as increased the proportion of
those having savings. It has also made program clients busier with larger number of
enterprises engaged in and more workers employed in these enterprises. No significant
impact, however, was found on assets and human capital investments.
The foregoing results led the authors to recommend that for microfinance programs to
be effective as a poverty-alleviation tool there is a need to review and constantly monitor
the effectiveness of the targeting procedures. In addition, it was pointed out that there
maybe a need to assist the poor in selecting appropriate projects that not only ensure
loan repayment but also generate ample profit as well.
Keywords: Microfinance, impact evaluation, quasi-experimental design, Philippines
IMPACT OF MICROFINANCE ON RURAL HOUSEHOLDS

IN THE PHILIPPINES
Toshio Kondo, Aniceto Orbeta, Jr., Clarence Dingcong, and Christine Infantado*
I.
INTRODUCTION
This paper reports on the impact evaluation study of the Rural Microenterprise Finance Project
(RMFP) in the Philippines. RMFP aimed to support efforts of the Government of the Philippines
to strengthen rural financial institutions by assisting organizations that employed the Grameen
Bank Approach (GB) in providing credit to the poor. The objective of the project was to reduce
poverty, create employment opportunities, and enhance the incomes of the poorest of the rural
poor (the ultra poor) the bottom 30% of the rural population as measured by income. Rural
banks, cooperative rural banks, cooperatives, thrift banks and non-government organizations
(NGOs) participated in the nationwide implementation of the project, which ended in December
2002.
II.
II.1
FRAMEWORK, SURVEY DESIGN AND DATA, ESTIMATION PROCEDURE

Framework for Assessing Impact on Households
The key problem in evaluation is finding a valid counterfactual against which the treatment
group is compared. The gold standard in impact evaluation is a randomized experiment where
treatment and control groups are randomly determined. Since this study was conducted ex-post,
i.e. evaluating an already completed project it could not employ a randomized experiment.
Moreover, the study also did not have the benefit of baseline data. Hence, the evaluation uses a
one-time survey, employing a quasi-experimental pipeline design as used by Coleman (1999) in
his study of microfinance in Thailand. The design is summarized in Table 1. Each treatment
barangay (village) is matched to a different comparison barangay.1 The importance of having a
different barangay rather than say a new center in the same barangay had been explained in
Coleman (1999). The treatment barangays are those where the Grameen Bank Approach
Replicators (GBAR) program, particularly lending, have been going on for some time. The
comparison barangays, on the other hand, are expansion areas where program clients have
been identified and organized into groups but no loans have yet been released to them. In both
the treatment and comparison barangays an equivalent number of qualified but nonparticipating households were also interviewed.
Toshio Kondo, Team Leader and Senior Evaluation Specialist, Asian Development Bank; Aniceto C. Orbeta, Jr.,
Econometrician/Impact Evaluation Specialist/Consultant of the Project and Senior Research Fellow, Philippine
Institute for Development Studies; Clarence Dingcong, Microfinance Specialist/Consultant of the Project; and
Christine Infantado, Portfolio Evaluation Officer, Asian Development Bank. Bihn Nguyen of the Economics and
Research Department, Asian Development Bank, provided assistance at the initial stage of the study. This report
also benefited from the extensive comments of David Levine, Professor, University of California, Berkeley. The
analyses herein, however, are the sole responsibility of the authors and do not reflect the views of the Institutions
they are affiliated with. Filipinas Gerardo and Paul Casuga assisted in supervising the nationwide field survey with
the generous cooperation of the PCFC and numerous officers of the respondent MFIs. This version of the report
also appeared as Network of Networks on Impact Evaluation (NONIE) Discussion Paper No 3, World Bank. This is
part of a larger study funded by the Asian Development Bank. The full report of the study is available at
http://www.adb.org/Documents/SES/REG/SST-REG-2007-19/SST-REG-2007-19.asp.
A barangay is a village, and is the smallest political unit in Philippines.
2
The innovation introduced in the study, not used by Coleman, is the inclusion in the group of
client households former clients consisting of graduates and problem households. This was
designed to address the attrition/drop-out problem in using new clients as comparison group,
i.e., the new client group included would-be graduates and future problem clients (Karlan,
2001).
Table 1: Evaluation Strategy: Type of Household Respondent
Type of households (HH) / Treatment (Existing) Area
Comparison (Expansion)
Area
Area
Participating HH
(A1) Existing clients
(C) New clients
(A2) Former clients
(graduates; problem clients)
Non-participating HH
(B) Qualified non-participating
(D) Qualified non-participating
From Table 1, impact is given by the expression:

(1)
Impact = (A-B)-(C-D).
This is also known in the literature as the difference-in-difference (DID) method. To see how the
DID method generate a clean measure of impact the cells in Table 1 can be filled by the factors
that determine outcomes for each of the different household clients. This is shown in Table 2.2
Table 2: Evaluation Strategy: Factors Determining Outcomes
Type
of Treatment (Existing) Area
Comparison (Expansion) Area
households (HH) /
Area
Participating HH
(A)
(C)
Observable characteristics
Unobservable
characteristics
Unobservable
characteristics
affecting participation
affecting participation
Area attributes (T)
Area attributes (C)
Microfinance program
Non-participating HH (B)
(D)
Area attributes (T)
Area attributes (C)
The new clients will not have the impact of the microfinance program because, even if they have
already been identified as prospective clients, they have not yet received loans. Nonparticipating households will neither have the effect of unobservable characteristics affecting
participation nor the impact of the microfinance program because they have not participated in
the program. A process of elimination will give the explanation why the DID method described
earlier will give the desired estimate of the impact of the microfinance program. The expression
(A-B) will give the net effects of unobserved characteristics affecting participation plus the
microfinance impact. Incidentally, this also highlights the effect of not controlling for sample
selection bias. The expression (C-D), on the other hand, will give the net effect of the
unobserved characteristics affecting participation. Thus, (A-B)-(C-D) will yield the net effect of
the microfinance program. It is note worthy that if we dont enumerate non-participating
households and compare say existing and new clients, (A-C) will give us the effect of the
microfinance program plus the difference between the treatment area and comparison area
effects which need not be identical, particularly if program placement is not random. Finally, if
2
The identified factors are adopted from de Aghion and Morduch (2005).
3
the treatment group does not include the appropriate number of former clients (graduates and
problem) the impact of both the observable characteristics and the unobservable characteristics
will be different for the existing and new clients as well. This is called the attrition/drop-out bias.
The DID strategy described above is implemented in a regression framework. The advantage of
using the regression framework is that it can account for the differences in household and
community characteristics which can happen even with a well-designed sampling scheme in a
quasi-experimental design. Specifically, the following equation was estimated:
(2)
Yij = F ( 1 X ij + 2V j + 3Mij + 4Tij + ij )
where:
Yij = household outcome of interest
Xij = household characteristics
Vj = village characteristics
Mij = membership dummy; 1 if participant in existing and expansion areas; 0 otherwise
Tij = treatment variable; 1 (or >0) if participant in treatment area3
The F() function can be linear or non-linear depending on the nature of the dependent variable
of interest. This expression is identical to the formulation in Coleman (1999) and Montgomery
(2005) had employed nearly identical evaluation strategy. As argued in Coleman (1999) and de
Aghion and Morduch (2005), conditional on the other regressors, the coefficient of Tij (4)
measures the impact of microfinance operations on household outcomes Yij. Woodridge (2002)
provides a discussion on the assumptions required for this result to extend to specific non-linear
cases like such as binary and corner solution responses.
This specification covers the three known sources of bias in evaluating the impact of
microfinance services using new members as comparison group. Control for non-random
program participation or sample selection is provided by using membership dummy M
(Coleman, 1999). The literature (e.g. Coleman (1999) and Armendariz de Aghion and Morduch
(2005)) has amply shown that not controlling for sample selection results in biased estimates of
the impact of microfinance services. Non-random program placement, on the other hand, is
controlled by village characteristics Vj or fixed effects estimation (Khandker, 1998). Finally,
dropout bias is controlled for by including in the treatment group an appropriate number of
randomly selected households who had dropped out of the program (both for reasons of
graduation and problems with repayments) as recommended in Karlan (2001).
II.2
The Survey
The survey requires two types of areas. First, treatment or existing areas defined as areas
where the program, particularly lending, has been ongoing for some time. In particular, existing
clients considered for the survey are those who have been with the program for at least 3 years
or have availed of loans for at least 5 loan cycles. This is designed to capture the impact of the
subject project, i.e., the RMFP, the implementation of which was completed in 2002. Second, a
corresponding set of expansion areas defined as areas where prospective program clients have
been identified and organized into groups but no loans have yet been released to them. A
3
This has also been rendered as
Yij = F ( 1 * X ij + 2 *V j + 3 M ij + 4 * M ij * Tij + ij )
variable; 1 (or >0) for treatment areas (cf. de Aghion and Morduch, 2005).
where Tij=treatment
4
suitable expansion area should be one that is different from an existing area. In particular, a
new center in a treatment area does not qualify as an expansion area.
The sampling design utilized the implementation structure of the RMFP. Participating MFIs
submit regular reports to the executing agency (EA) - the Peoples Credit and Finance
Corporation (PCFC). The records of the PCFC provide the number of clients actually served by
each MFI at the barangay level. There was no comprehensive record on expansion areas.
While most of the participating MFIs claimed to have expansion areas, a check with a few of the
MFIs, however, revealed that some MFIs did not have the suitable expansion areas needed by
the study. The sampling then used the list of existing barangays served as the sampling frame.
The sampling scheme considered the three island groups (Luzon, Visayas, Mindanao) and the
type of MFI (cooperative banks/rural banks, cooperatives, and NGOs) as stratification variables.
Based on existing program records, it was determined that the most practical primary sampling
unit (PSU) is the barangay. It was also determined based on the estimates of the mean and
variance of incomes from the Family Income and Expenditures Survey (FIES) in 2003 that a
sample size of 2,200 households was sufficient for the study. For each barangay a sample of 10
client and 10 non-participating households were deemed sufficient. At this sampling rate per
barangay, about 110 barangays or 55 treatment (existing) barangays and 55 corresponding
comparison (expansion) barangays will be required to generate the needed sample size.
The number of barangays for each island and for each MFI type is selected randomly
proportional to the number of client households served or sampling proportional to size (PPS).
For every treatment barangay selected, the MFI concerned is asked to identify a corresponding
suitable expansion area. The selection of a particular treatment barangay for inclusion in the
survey is contingent on the MFI being able to identify a corresponding suitable expansion
barangay. When the MFI cannot identify a suitable expansion barangay, the treatment barangay
is replaced with a new draw from the pool of treatment barangays for the same MFI type. This
process is repeated until the required number of treatment-expansion barangay pairs are
generated for each of the MFI types. The existing and new client households are drawn
randomly from the list prepared by the MFI or from the centers roster of members. The nonparticipating households are drawn randomly from the list of qualified non-participating
households identified by MFI field personnel, center or barangay leaders.4
The total number of borrowers by island group and MFI type as of 30 June 2006 is given in
Table 3. The corresponding allocation of the treatment barangays by island group and by MFI
type is given in also given in the table. The survey covered 2,276 households in 116 barangays
and 28 MFIs.
The MFIs did not keep a record of eligible households in the communities they are operating. Thus there is no way
of knowing whether the list of non-participating household is comprehensive or not without going into a listing
operation. Household listing, however, was not done due to resource limitations. This should be considered as a
limitation of the study.
5
Table 3: Sampling Allocation by Island Group and Type of MFI
Island Group
Grand Total
Luzon
Banks
Coops
NGOs
Visayas
Banks
Coops
NGOs
Mindanao
Banks
Coops
NGOs
Total
Banks
Coops
NGO
Total No. of Borrowers

1,648,052
797,194
485,984
70,461
240,749
419,123
67,125
69,046
282,952
431,735
331,097
41,331
59,307
1,648,052
884,206
180,838
583,008
%
100
48
61
9
30
25
16
16
68
26
77
10
14
100
54
11
35
Treatment Barangays
55
28
18
2
8
13
2
2
9
14
10
2
2
55
30
6
19
Source: PCFC
Three survey instruments were used in the study. One is the household survey questionnaire.
The questionnaire is adopted from the Annual Poverty Indicators Survey (APIS) questionnaire
conducted by the National Statistics Office. Added to the APIS questions are the detailed
questions on loan accounts, enterprises and gender-related matters. Another instrument is the
Barangay Profile Questionnaire. Finally, there is the MFI Profile questionnaire. These
instruments were pre-tested prior to the actual field surveys.
II.3
Estimation Procedure
The estimation methodology considers the nature of the dependent variable and the treatment
variable. It follows closely the estimation procedures described in Wooldridge (2002) for
estimating the average treatment effects. Before discussing the estimation procedures, it is
useful to discuss the nature of the treatment variables and the outcome variables considered in
the study.
Outcome variables: Several outcome variables are considered in the study, namely: (a) basic
household welfare measures such as per capita income, per capita expenditures, per capita
savings, and food expenditures; (b) other financial transactions such as other (non-GBA) loans
and personal savings stocks5; (c) household enterprises and employment; (d) household assets
such as land, farm equipment, livestock and poultry, and household appliances; and (e) human
capital investments such as education and health. Some of these variables are continuous such
as per capita income, expenditure, savings, food expenditure, health expenditure per capita,
and education expenditure per attending child. Others are binary such as having a savings
account and availing of a non-GBA loan. Others are truncated such as value of household
assets and other loans. Others are count variables such as the number of non-GBA loans,
number of enterprises, and the number employed in those enterprises. Finally, others are
proportions such as the proportion of school-age children attending school or proportion of
those who are sick to sought treatment. Each of these different types of dependent variable
requires different estimation methodology.
This refers to savings (stocks) accounts held by the respondent in the program MFI or other MFIs and is different
from the savings (flow) variables measured as the difference in income and expenditures.
6
Treatment variables: There are four possible treatment variables that can be used to assess the
impact of microfinance on household welfare. These are: (1) availed program loan (1=yes,
0=otherwise); (2) number of months the program is available to the barangay (based on first
loan released for the barangay); (3) value of loans (cumulative total amount of loans) availed
and (4) number of loan cycles. The length of exposure to the program is expected to have an
impact. Therefore treatment variables (2)-(4) are deemed better in representing program
availability (Coleman, 1999). It should be realized, however, that these treatment variables have
different implications for estimation. For instance, perhaps only the first two satisfy the
ignorability of treatment6 condition for treatment variables. Treatment variables (3) and (4) would
fail the ignorability condition and would thus require instrumental variable estimation
(Wooldridge, 2002).
Other Independent variables. The other independent variables used in the control functions are
similar to those used in existing literature (e.g. Coleman 1999, Montgomery 2005). These
include household characteristics such as age of the reference person (a.k.a. household head)
or respondent; education of the reference person7; number of years in the barangay, and house
size. Age is expected to be a factor because it is well-known that age-earning profile is not flat.
Education, of course, is a known determinant of both earning capacity and productivity in nonmarket (home) production. The number of years in the barangay is a proxy for social capital.
House size is a proxy for household wealth8. This is used because, among the household asset
in the data, this is presumed to be the least volatile. For education and health equations, the
variables indicating availability of relevant facilities are also added as explanatory variables.
Estimation method: The general estimation methodology can be labeled as control function
approach. This approach uses other independent variables as elements of some control
function in addition to the treatment variable. The functional form of the control function depends
on whether the outcome of interest can be modeled linearly or not. For outcomes that can be
modeled linearly (i.e., y=x) such as continuous variables, the elements of the control function
include the other independent variables, such as household characteristics, and the interaction
of the treatment variable and the demeaned values of the other independent variables. For
linear models, the coefficient of the treatment variable provides the estimate of the DID average
treatment effect. For outcomes that require non-linear models (i.e. y=F(x)) such as probit for
binary outcomes, tobit for truncated outcomes or poisson for count outcomes, Wooldridge
(2002) recommends that propensity score method is more appropriate. Under this method the
propensity score, which is the predicted value of the regression of the treatment variable on the
other independent variables, and the product the treatment variable and the demeaned values
of the estimated propensity score are the elements in the control function. In non-linear models,
the DID average treatment effect is given by the marginal effects of the treatment variable
measured at the average values of the independent variables subject of course to satisfying the
required assumptions. Note that the correction for sample selection is taken cared of by the
inclusion of membership dummy among the explanatory variables. To take care of non-random
program placement, fixed effects estimation is used9. However, in general, fixed effects
6
7
8
Originally attributed to Rosenbaum and Rubin (1983). This is defined as conditional on observable characteristics,
the treatment and outcome variables are independent. Practically, it means that the treatment variable must not be
under the control of or exogenous to the respondent.
Coleman (1999) prefers to use the highest educational attainment achieved by any member in the household.
The ideal wealth variable would be household assets pre-dating the availability of the program. This was not
available from the data set because of recall problems. Coleman (1999), for instance, used value of asset acquired
five years ago.
Barangay variables could have been used, but the study experienced significant refusals for the Barangay Profile
Survey (BPS) which would significantly reduce the number of observations if used. Eleven barangays did not
7
estimation will result in inconsistent estimates when a non-linear model is estimated
(Wooldridge, 2002). Thus for these models, random effects estimation were used. Admittedly,
random effects estimation is more restrictive than fixed effects because it imposes a structure
on the village effects. This is, however, considered better than the inconsistent estimates from
fixed effects estimation with non-linear models. Finally, as mentioned earlier, the nature of the
treatment variable also determines the estimation procedure. When the ignorability of treatment
cannot be assumed (such as those for treatment variable (3) and (4)) instrumental variables (IV)
estimation is used (Wooldridge, 2002). For lack of better instruments, we will use treatment
variable (2)10 as instruments for all estimations using (3) and (4). The validity of the treatment
variable (2) as an instrument emanate from the fact that whatever loans existing clients are able
to get as well the number of loans cycles are all dependent on the number of months the
program is available in the area. In addition, this variable is determined by the MFI and not
within the control of the households.
III.
ESTIMATION RESULTS
III.1
Respondent Characteristics and Outreach
When RMFP was completed in 31 December 2002, 618,906 clients were reached, of which
97% were women. When the survey design was being formulated in June 2006, records of the
PCFC showed that program had served some 1.6 million borrowers. The survey showed that
existing clients are, on average, on their 75th month in the program or about 6 years and 3
months (Table 4). They have, on average, cumulatively availed of some P70 thousand in loans
and they are on their 7th loan cycle. It is also revealed that, on average, 9% of exiting clients are
problem clients while 2% are graduates.
Table 4: Characteristics of Existing Clients
Characteristics
Months since first loan
Total amount of loans, thousands
No. of loan cycles
Existing, %
Graduate, %
Problem, %
Values
75.2
69.923
7.2
89.1
2.1
8.8
Source: Operations Evaluation Mission
Table 5 shows the demographic characteristics of all respondent households. It shows that the
respondents are 44 years old on average. Fifteen percent of the reference persons11 are
female. In terms of education, below 1% have no education, 31% have some elementary
accomplish the BPS which would mean removal of about 220 household respondents if the barangay profile data is
used.
10
Other candidate variables would be barangay characteristics. However, as mentioned earlier, the study
experienced significant refusal problem with this instrument. We are grateful to D. Levine for pointing out that using
treatment variable (2) as instrument for treatment variables (3) and (4) will not be very different from using
treatment variable (2) directly. Since these treatment variables did not turn out to be significant in the estimation
results these were not used in subsequent discussions.
11
Reference person is the person in the household with whom all relationships with other household members
referenced with. This person is commonly known as the household head.
8
education, 46% have some secondary education and 23% have tertiary education. The
respondents have lived in the barangay for about 19 years and the average size of the house is
63 square meters. About 92% of the respondents are female. Since for existing clients the
respondents are the program clients, one can say that program clients are 95% female. Table 6
shows the basic household welfare indicators. Using the official poverty threshold in 200612, the
table also shows that only 10% of the respondents are poor while 4% are subsistence poor13.
This is quite revealing considering that the program was designed to reach poor households.
Table 5: Demographic Characteristics of Respondents
Existing Areas
Participa- Non-Parti- Sig.
ting
cipating
Variables
Age of reference person (rp)

Female, rp
Less than elem, rp
Elementary, rp
Secondary, rp
Tertiary, rp
Years in barangay
House size, sq. m.
47
0.146
0.008
0.337
0.442
0.214
21.2
75.7
43
0.140
0.009
0.273
0.452
0.266
17.6
65.9
Female, respondent
0.953
0.926
***
*
*
***
*
Expansion Areas
Participa- Non-Parti- Sig.
ting
cipating
Total
43
0.127
0.009
0.293
0.484
0.214
17.6
59.4
44
0.197 ***
0.004
0.347
0.442
0.207
18.0
51.2 **
44
0.153
0.007
0.312
0.455
0.225
19
63
0.905
0.911
0.923
***, **, * - significant at 1%, 5% and 10%, respectively; two-tailed hypothesis

Table 6: Basic Welfare Indicators of Respondents
Variables
Per capita income
Per capita expenditure
Per capita savings 1
Per capita savings 2
Per capita food exp.
Poor \a
Subsistence poor \b
Existing Areas
Participa- Non-Partiting
cipating
51,000
45,365
36,153
34,357
14,847
11,007
18,425
14,508
13,708
13,115
0.060
0.025
Expansion Areas
Sig. Participa- Non-Parti- Sig.
ting
cipating
*
43,737
43,456
30,674
31,898
**
13,064
11,558
**
15,454
14,210
12,540
13,145
0.110 ***
0.042
0.093
0.032
0.120
0.054
Total
45,759
33,195
12,564
15,580
13,113
0.097
0.039

Per capita savings 2 recognizes that the benefits of expenditure on education, health and durable furniture extends
beyond the current reference period. Source: Operations Evaluation Mission
Given the intention of the program to serve poor households, the natural question to ask is is
the program reaching its intended target? To answer this question the distribution of the
difference between the respondents per capita income and the official poverty threshold was
plotted. As such, a value of zero would mean the household is on the poverty line, a negative
value would mean the household is below the poverty line and a positive value would mean the
household is above the poverty line. Figures 1-3 show the histograms of the deviation of per
12
This is published in the National Statistics Coordination Board website (http://www.nscb.gov.ph). The national
poverty threshold for rural areas is estimated to be 13,659 while the food threshold is 9,445 as of 7 March 2006.
13
Defined as those below the food threshold.
9
capita income from the poverty threshold for existing clients, new clients, and non-participating
households. The histograms reveal that while a large proportion of the respondents are around
the poverty threshold more are on the non-poor side.
If one considers only existing clients, the histogram shows that the while a considerable
proportion of the existing clients are around the poverty threshold, a larger proportion is on the
non-poor side (Figure 1). One can argue that perhaps when they entered the program they were
poorer than they are at the time of the survey. If one, however, looks at the profile of the new
clients, one sees essentially the same distribution, i.e., larger proportions are on the non-poor
side (Figure 2). Finally, the non-participating households, which are households people in the
community considers as qualified for the program, also exhibit the same characteristics, i.e.
most of them are not poor (Figure 3). These graphs reveal some very important pieces of
information about the program. One, since existing and new program clients are supposed to
have been screened using means-testing procedures, these procedures assuming they have
been applied strictly, are not correctly identifying the poor clients per official definition. Two,
since non-participating households are households referred to by either program field
personnel, center leaders or barangay leaders as those who would qualify for the program,
these stakeholders in the field are also pointing at possible clients that are not the intended
clients of the program. These means that, in spite of the means-testing instruments used to
identify the intended clients, all relevant stakeholders of the program in the field are not pointing
to the intended clients the poor as the qualified clients of the program. This may indicate that
perhaps the stakeholders are conveying the message that those considered as officially poor
may not be the desired clients of microfinance programs.
Figure 1: Distribution of Sample Existing Client Household Around the Poverty Line
p
200,000
1 0 0 ,0 0 0
20,000
0 = t h re s h o ld
.1
F ra ct io n
.2
.3
.4
p
E x is t in g C lie n t s
D e v i a t io n o f in c o m e f ro m p o v e rt y th r e sh o ld
Source: Operations Evaluation Mission.
10
Figure 2: Distribution of Sample New Client Households Around the Poverty Line
2 00,00 0
1 00,00 0
0 = t h r e s h o ld
20,00 0
F ra c ti o n
.1
.2
.3
N e w C lie n t s
D e v ia ti o n o f i n c o m e f r o m p o v e rt y t h r e s h o l d
Figure 3: Distribution of Sample Qualified Nonparticipating Households

Around the Poverty Line
F ra c t io n
.1
.2
.3
D i s t r ib u t i o n o f s a m p l e h o u s e h o l d s a r o u n d p o v e r t y l i n e
Q u a l i f i e d N o n - P a r ti c i p a t i n g
200,0 00
1 00,000
2 0,000
0 = t h r e s h o ld
D e v ia t io n o f in c o m e f r o m
p o v e r t y th r e s h o l d
III.2
Estimation Results - Impact of Microfinance
Impact on per capita income, expenditure, savings, and expenditure on food

The primary measures of household welfare are, of course, per capita income, expenditure,
total and food, as well as savings. These variables are continuous, hence, the estimation
procedures uses linear fixed-effects model. The control function variables include other
independent variables, such as household characteristics, and the interaction of the treatment
variable with the other independent variables (expressed as mean deviations). As mentioned in
Section III.3, four treatment variables can be used in the study, namely: (1) availed program
loan (1=yes, 0=otherwise); (2) number of months the program is available to the barangay
(based on first loan released for the barangay); (3) value of loans (cumulative total amount of
loans) availed and (4) number of loan cycles. Estimation results show that among the four only
the availed program loan treatment variable turned out significant and this is only true for per
capita income, per capita consumption and per capita food consumption while it is insignificant
11
for the two savings definitions. The F-test on whether the fixed-effects coefficients are all equal
zero is rejected, which lends support to the hypothesis of non-random program placement. For
detailed estimation results please see Annex Table A1-A5.
Table 7 shows the summary of the impact of availing of program loans on per capita income,
per capita expenditures, on two definitions of per capita savings and per capita food
expenditures. The table shows a mildly statistically significant (significance level 10%) positive
impact on per capita income of the availed program loan treatment variable. The other treatment
variables did not show significance even at 10% level. The estimated parameter says availing of
program loan means higher income by about P5,222 compared to those who have not availed
of the loans. Translating this into impact on per loan availed requires some calculation. The
dependent variable is average annual per capita income. Considering that on average
households have availed a cumulative P70 thousand in loans in 6 years or about P11,000 per
year, this means that every P100 loan availed income increase by P47.
Table 7, also shows that per capita expenditure is also positively affected by access to program
loans. The estimate puts this at about 4,136 pesos. Using the same calculation employed
earlier this would mean 38 pesos increase in per capita consumption per every 100 pesos loan
availed.
These estimates on the impact on income and consumption are higher than the well known
estimate of 18% percent (Pitt and Khandker 1998) and 10% (Khandker, 2003). Of course, both
used consumption as the dependent variable. Zeller et al. (2001) using data also from
Bangladesh generated similar higher estimate of the impact - an annual average of about 37% with per capita income as the dependent variable. They used access to credit, rather than loans,
as the treatment variable like in this study. They have explained the difference by noting that in
using access to credit rather than loans they have captured not only the benefits of loans but
also other indirect benefits from the ability to borrow if needed which would include, for instance,
reduced cost of consumption smoothing such as decrease in distress sale and increase riskbearing capacity favoring more profitable production and investment portfolios. Of course, one
needs to add that the precision of the estimates is lower in this study.
Savings, in its two definitions14, is not significantly affected. Finally, per capita expenditure on
food is also positively affected. The estimate is about P1,333 higher compared to those who did
not avail of program loans or about P12 per every P100 peso loan availed.
Table 7: Impact on per capita income, expenditures, savings, food
Outcome variables
Per capita income
Per capita expenditures
Per capita savings 1\a
Per capita savings 2\b
Per capital food exp
Estimated
Coefficient
5,222
4,136
1,333
Sig. Level
0.099
0.077
ns
ns
0.072
a=Income-Expenditure
b=Income-Expenditure+Educ+Health+Dur. Furniture
14
One definition is income minus expenditures. The second definition adds back expenditures on education, health,
and durable furniture because these are not expected to be consumed in one period (see for instance Bautista and
Lamberte, 1990).
12
Since only the loan availability/access treatment variable was found to be statistically significant
on primary measures of welfare, such per capita income and expenditure, subsequent
discussions will be limited to this treatment variable.
Impact on Other Loans and Personal Savings
Besides household income and expenditures, it is also important to look at the impact of the
program on the other financial transactions of the household. Among the important financial
transaction of household are other loans and maintaining savings account of program clients.
The savings referred to here are accounts maintained in the program MFI and other MFIs and
thus can be considered stock rather than flow savings variable discussed earlier.
About one-fourth of the respondents have availed of non-GBA loans in the last two years (Table
8). About 20% of existing clients have availed of non-GBA loans while a higher proportion (26%)
of new clients and non-participating household have availed of non-GBA loans. The amount of
non-GBA loans, however, is higher for existing clients (P20 thousand) compared to new clients
(P9 thousand) and non-participating households (P17 and 12 thousand). In terms of the number
of loans contracted, the existing clients have a higher number (1.6) compared to new (1.2) and
non-participating households (1.2).
Table 8: Non-GBA loans
Variables
Existing Areas
Expansion Areas
Participa- Non-Parti- Sig. Participa- Non-Parti- Sig.
ting
cipating
ting
cipating
0.201
0.261 **
0.258
0.268
Availed non-GBA loans

Among those with non-GBA loans:
Amount of other loans, thousands
20.335
17.754
8.776
Number of other loans
1.648
1.263 ***
1.179
12.357
1.151
Total
0.248
14.328
1.280
To determine the impact of the program on the financial transactions of households, models are
estimated on three variables: (a) availed non-GBA loans; (b) amount of these other loans; and
(c) number of loans transacted. The availed non-GBA loan decision was estimated as a binary
choice using the Probit model. The amount of non-GBA loans decision was estimated as a
truncated variable using Tobit model. Finally, the number of non-GBA loans transacted was
estimated as a count variable using Poisson regression. All of these are non-linear models.
Thus, as mentioned in III.3, the control function uses the propensity score method. The results
show that the treatment variable - availed of program loan is significant in the availed nonGBA loan variable and insignificant in the amount and number of non-GBA loans contracted.
Details of the estimation results are provided in Table A6-A7.
Table 9 provides the summary of the estimates on the impact on non-GBA loans. The
estimates show that availability of program loans significantly, albeit mildly (significance 6
percent), reduced the use of non-GBA loans. The estimated coefficient shows that compared to
non-program respondents the non-GBA loans contracted in the last two years was reduced by
about 5%15. As mentioned earlier, in terms of loan amount and the number of non-GBA loans
contracted, however, the impact is statistically insignificant.
15
This used a non-linear probit model so the coefficient does not provide the marginal effects on the probability of
contracting an non-GBA loan of availing of program loans.
13
Table 9: Impact on non-GBA loans

Marginal
Effects
-0.0530
Availed non-GBA loans

Amount of non-GBA loans
Number of non-GBA loans
Sig.
Level
0.056
ns
ns
In terms of personal savings16 (stocks), about two-thirds of the respondents maintain personal
savings accounts either in the program or non-program MFIs (Table 10). This is higher for
existing clients (86%) compared to the new clients (66%) and non-participating (54% and 52%)
households. In terms of balances, about 70% of household have 1 to 5,000 pesos, 15% have
5,000 to 10,000 pesos and the proportion with more than 10,000 pesos.
Table 10: Saving accounts in program and other MFIs
Variables
Have personal savings account
Among those with personal savings:
Personal savings 1-5,000
Personal savings 5,000-10,000
Personal savings 10,000+
Existing Areas
Expansion Areas
ting
cipating
ting
cipating
0.859
0.535 ***
0.657
0.521 ***
0.653
0.181
0.166
0.652
0.154
0.194
0.794
0.128
0.078
0.663
0.139
0.198 ***
Total
0.637
0.697
0.149
0.154

To determine the impact of the program on personal savings, models on decision to maintain a
savings account and the balances on those accounts were estimated. The decision to maintain
a savings account is estimated as a Probit model while the modeling for the savings balances
uses an Ordered Probit model considering the three ordered categories of savings balances
mentioned earlier. Again these are non-linear models so propensity score method was used for
the control function. The estimation results are given in Table A8. The results show that both
maintaining personal savings accounts and the amount of balances are significantly affected by
availing of program loans.
Table 11 shows the summary of the estimation results indicating that the impact of the program
is positive and highly statistically significant (significance less than 1 percent) both in terms of
having a personal savings account as well as on the amount of savings. The estimates shows
that compared to those who are not program clients, as much as 23% more of those who have
availed of program loans have maintained savings account. In terms of the amount of savings,
those with 0 to 5 thousand is lower by 12%, those will 5-10 thousand is higher by 4% and those
with 10 thousand or more is higher by 9% compared to those who have not availed of program
loans.
16
The respondents were asked about their savings accounts both in program MFI as well as other MFIs.
14
Table 11: Impact on savings in program and other MFIs
Have personal savings
Amount of personal savings
0 - 5,000
5,000 -10,000
above 10,000
Marginal
Effects
0.230
Sig.
Level
0.000
-0.124
0.038
0.086
0.003
0.001
0.005
Impact on the Number of Enterprises and Employment

Another important impact of microfinance is what it does to the enterprises the respondent
households are engaged in. The survey asked respondents about the enterprises and
employment in these enterprises of program clients as well as other household members.
Table 12 shows that about 93% of the existing client respondents have household enterprises.
This is higher compared to new clients (87%) and non-participating households (78%). Among
those with household enterprises, the number of enterprises is about 1.8 and the total number
of employed people is about 2.4. For existing clients this is about 2.1 enterprises employing
about 3 individuals, for new clients this about 1.8 enterprises employing 2.4 individuals and for
non-participating households in treatment and expansion areas this is 1.6 enterprises employing
2.4 and 2.0 individuals, respectively.
Table 12: Household enterprises and employment
Variables
Existing Areas
Expansion Areas
ting
cipating
ting
cipating
0.926
0.777 ***
0.871
0.779 ***
With household enterprise

Among those with household ent.:
Total number of enterprises
2.07
1.63 ***
1.82
Employed family members
2.31
1.62 ***
1.68
Employed non-family members
0.63
0.78
0.68
Total employed
2.95
2.40 **
2.36
1.63 ***
1.66
0.33
2.00
Total
0.836
1.79
1.82
0.61
2.43
Since substitutions can happen between program clients enterprise and those of other
household members, the analysis is focused only on total enterprise and total employment. The
modeling of the number of enterprises and number of employees considered these variables as
counts and were estimated using Poison regression. Since this is a non-linear model, propensity
score method was used for the control function. The estimation results show that the availing of
program loans significantly affect the number of enterprises households have as well as the
number employees in these enterprises. Details of the estimation results are provided in Tables
A9-A10.
The estimates show that the impact of the program on both the number of enterprises as well as
the number of employed persons in these enterprises is a very significant positive. Table 13
shows that compared to non-program households, the number of enterprise households with
15
program clients is higher by 20%17. The table also shows that households with program clients
have 17% more employed person than non-program clients.
Table 13: Impact on enterprises and employment
Total number of enterprises

Total number of employees
Incidence
Rate
1.20
1.17
Sig
Level
0.009
0.006
Impact on Assets
It is very likely that microfinance will affect the acquisition of assets of households. The
respondent households were asked about the current value of the assets they have. The
standard question asked about assets is if someone wants to buy a particular asset owned how
much would the price be? The assets include land, equipments, livestock and poultry and
household amenities.
Table 14 shows that about 20% of the respondents have land assets with an average current
value of P557 thousand. It also shows about 15% of the respondents own farm equipment with
an average current value of P55 thousand. About 53% of the respondents have livestock and
poultry assets with an average current value of 46 thousand. Finally, almost everyone (97%)
have some household appliances with current value of 73 thousand.
The value of total assets household was estimated as a truncated variable (having only positive
values) using Tobit model. In fact, as Table 14 shows, substantial proportion of households
does not have specific assets. Since this is a non-linear model, propensity score method was
used for the control function. The estimations did not show significant impact of the program in
the total household assets. Details of the estimation results are provided in Tables A11-A12.
Table 14: Household assets
Variables
Existing Areas
Expansion Areas
ting
cipating
ting
cipating
0.198
0.179
0.231
0.190
With agriculture and comm land

Among those with ag and comm land:
Agri and commercial land, cur. value
468,338 581,081
580,559
With farm equipment
0.117
0.124
0.194
Among those with farm equipment:
Farm equipment, cur. value
27,588
34,108
110,645
With livestock and poultry
0.586
0.490 ***
0.565
Among those with livestock and poult.:
Livestock and poultry, cur. value
20,419
18,556
83,484
With household appliances
0.981
0.973
0.968
Among those with household appl.:
Household appliances, cur. value
59,547
66,689
121,606
590,688
0.167
Total
0.199
557,332
0.151
24,885
0.472 ***
55,365
0.527
58,183
0.958
46,028
0.970
44,252
73,311
17
Since this is a Poisson regression, incidence rate (exp(coefficient)) is given. This is given in the column labeled as
IRR.
16
Impact on human capital investments (education and health)

Any change in income or expenditure does not necessarily translate into changes in human
capital investments. The education variables examined are school attendance proportion for
school-age children (6-12 years, 13-16 years, and 17-24 years) as well as the education
expenditure per school attending child. For health, the variables examined are the proportion of
household members who are ill or injured, proportion of those ill or injured who sought medical
treatment, proportion of children 0 to 5 year who are fully immunized and per capita health
expenditures.
Table 15 shows that about 95 percent of children 6-12 years old, 87 percent of children 13-16
years, and 31 percent of children 17-24 years old are attending school. It also shows that the
average expenditure per attending school-age child is about 7,239 pesos. In terms of health
indicators, Table 16 shows that the proportion of either sick or injured in the past 6 months
preceding the survey is about 9 percent. The proportion of households with at least one
ill/injured member is about 23%. The proportion of those ill/injured who sought treatment is 69%.
The proportion of children 0-5 years old who are fully-immunized is about 69%. The average per
capita expenditure for health is about 740 pesos.
Table 15: Education outcomes
Variables
With children 6-12 years old

Among those with children 6-12:
Proportion attending school, 6-12
Educ exp per school age child
Educ exp per attending child
Existing Areas
Expansion Areas
ting
cipating
ting
cipating
0.502
0.499
0.576
0.472 ***
Total
0.512
0.970
0.416
0.959
0.345 **
0.944
0.393
0.927
0.326 **
0.950
0.369
0.881
0.490
0.915
0.372 ***
0.869
0.426
0.817
0.439
0.871
0.430
0.344
5,931
8,241
0.319
6,301
8,313
0.306
4,615
6,300
0.278
4,525
6,128
0.312
5,312
7,239

Table 16: Health outcomes

Variables
Existing Areas
Expansion Areas
ting
cipating
ting
cipating
0.097
0.072
0.101
0.091
0.255
0.181 ***
0.269
0.213 **
Proportion of members ill/injured

With illness/injured members
Among those ill/injured members:
Proportion who seek treatment
0.700
0.685
0.696
0.323
0.430 ***
0.444
Among those with children 0-5 years old:
Proportion fully-immunized
0.717
0.666
0.719
Per capita medical exp
645
954
560
Total
0.090
0.229
0.657
0.394
0.686
0.399
0.659
791
0.689
740
17
Two types of variables were considered in modeling the impact of the program on education
variables, namely: (a) proportion of school-age children attending school and (b) expenditure
per attending child. The proportion of school-age children attending school was used to in order
to be closer to the notion of a household decision variable.18 With a proportion as a dependent
variable, the fractional logit model19 was used to estimate it. In the case of expenditure per
attending child this was estimated using linear fixed-effects model because this is a continuous
variable. In terms of the control function, the fractional logit model, being a non-linear model,
used propensity score while the expenditure per attending child used the interaction between
the treatment variable the demeaned values of the other independent variables. The estimation
results show that availing of program loan is not significant for school attendance for all age
groups as well as for expenditure per attending child. Details of the estimations results are given
in Tables A14-A15.
In the case of health, four variables are used, namely: (a) proportion ill or injured, (b) proportion
who seek treatment if ill or injured, (c) proportion of fully immunized children 0-5 years old, and
(d) per capita medical expenditures. Similar to the treatment of the school attendance variables,
variables (a) to (c) were modeled as proportions using fractional logit model. Per capita medical
expenditures, on the other hand, was modeled as continuous variable using linear fixed-effects.
The results show that availing of program loan does not significantly affect for all health
variables under consideration. Details of the estimation results are provided in Tables A15-A16.
Impact on Hunger Incidence and Reduction in Food Consumption
Hunger incidence as well the reduction in food consumption in the last three months was
likewise studied as well. Table 17 shows that hunger incidence is about 2% in the respondent
population. The reduction in food consumption in the last three months, on the other hand, is
found in 11 percent of the respondent households.
Hunger incidence was modeled as binary outcome using the Probit model. Since this a nonlinear model, propensity score method was used for the control variables. The estimation results
show that availing of program loan did not significantly affect the incidence of hunger. Details of
the estimation results are provided in Tables A17-A18.
Table 17: Hunger and reduced food consumption incidence
Variables
Hunger incidence
Reduced food incidence
Existing Areas
Expansion Areas
ting
cipating
ting
cipating
0.023
0.032
0.019
0.014
0.113
0.124
0.110
0.105
Total
0.017
0.107
Impact by Different Socioeconomic Groups

The evaluation of the program was also designed to test whether the impact of access to
microcredit differed across socioeconomic groups. While the poor/nonpoor distinction is useful,
a better picture is given by dividing the sample households into per capita income quartiles.
18
This can also be estimated as an individual-based decision model, i.e. attendance of each school-age child is
treated independently. This, however, may not capture the idea that the attendance of all school-age children in the
household are jointly decided on by parents.
19
Adopting method used in Papke and Wooldridge (1996).
18
There are a couple of ways of estimating the impact on different subgroups. One is estimating a
separate equation for each subgroup. Another is to jointly estimate the impact in a single
equation using the interaction of subgroup and treatment variables, i.e., using the coefficient of
the interaction between the availability treatment variable and corresponding quartile dummy
variables to measure the impact for each quartile. Orr (1997) argues that the latter approach
has two advantages: (i) it usually provides more power because it uses the full sample to
estimate the coefficients; and (ii) it allows one to test whether there are statistically significant
differences in impact among the subgroups taken as a set (rather than between pairs of
subgroups). Given these considerations, the joint estimation method was adopted for this study.
The survey respondents were divided into four quartiles, i.e., those (i) with annual per capita
incomes less than P21,480; (ii) from P21,481 to P34,428; (iii) from P34,429 to P56,167; and (iv)
over P56,167. For comparative purposes, it is useful to mention that the poverty line in the
Philippines is equivalent to an annual per capita income of P14,405 and for rural households
this is P13,659.20
The results show that the program had a regressive impact (Table 18). A significant positive
impact was evident only for the households in the top quartile while there was a negative impact
on the poorer households. For instance, per capita income for the participating households in
the poorest quartile was P23,000 lower than the nonparticipating households. However, the
impact for the top quartile was positive and resulted in a P45,000 increase in annual income
compared with the nonparticipating households in the same income group. The results were
similar for the per capita expenditure, savings, and food expenditure. It is worth noting that the
impact on savings is significant for all except the third quartile in contrast to the insignificant
impact for the whole sample. Table A19 provides the details of the estimation results.
Table 18: Impact on Household Outcome by Per Capita Income
Quartile 1
Quartile 2
Quartile 3
Quartile 4
Per Capita Income

Per Capita Expenditure Per Capita Savings 1
Per Capita Savings 2 Per Capita Food Expenditures
Coeff.
Sig. Lev.
Coeff.
Sig. Lev.
Coeff.
Sig. Lev.
Coeff.
Sig. Lev.
Coeff.
Sig. Lev.
-23,214.0
0.000
-9,459.7
0.007 -13,754.2
0.000 -14,567.2
0.000
-3,476.9
0.002
-13,903.1
0.001
-6,752.6
0.034
-7,150.6
0.006
-7,680.6
0.005
-1,408.5
0.164
-1,212.2
0.764
1,849.6
0.548
-3,061.8
0.228
-3,010.6
0.251
1,382.1
0.159
45,113.7
0.000 23,915.6
0.000 21,198.1
0.000 23,928.6
0.000
6,659.9
0.000
Coeff. = Coefficients, Sig. Lev.. = Significance Level

There are several possible reasons explaining why the impact on the lower income households
is lower (or negative). These include: (i) the problem clients are concentrated among the poorer
households; (ii) the average size of loans may be smaller for poorer households; (iii) there may
be a preponderance of diversion of loan proceeds from production to consumption among
poorer households; and (iv) if there is no diversion, the projects of poorer households may be
less productive. In this study, there is empirical evidence only for (i) and (ii).
Indeed, the average loan size for poorer households is smaller (Table 19). This prevents them
for venturing into more productive activities that would require more capital. In Table 20, it is
shown that there are more problem clients among the bottom three quartiles. While the average
proportion of problems in the sample is about 2%, the bottom three quartiles have each a little
over 2% that are problem clients while the highest per capita income quartile group had only
less than 1% as problem clients.
20
Estimates as of 7 March 2006 from the National Statistics Coordination Board website. Available:
http://www.nscb.gov.ph
19
Table 19: Distribution of Average Loan Size by Per Capita Income Quartile
Quartile
Mean Cumulative
Loans
(P000)
45.031
57.540
64.290
99.168
69.923
Lowest
Lower middle
Upper middle
Highest
Total
Total
Number of
Cycles
6.1
6.9
7.1
8.2
7.2
Average Loan
Size
(P000)
7.392
8.280
9.087
12.166
9.721
Table 20: Distribution of Type of Clients by Per Capita Income Quartile (%)
Quartile
Lowest
Lower
Middle
Upper
Middle
Highest
Total
Existing
Graduates
Problem
New
14.96
18.43
0.55
0.54
2.37
2.68
27.01
26.30
NonParticipating
55.11
52.06
21.34
0.54
2.17
27.49
48.46
100
27.50
20.59
0.36
0.50
0.89
2.03
21.61
25.59
49.64
51.31
100
100
Total
100
100
The regressive relationship provides further evidence that microfinance projects should not
target the ultra poor. Additional debt may make their lives worse, not better. Coleman (2006),
using data from Thailand, qualified the earlier no significant impact on consumption result in
Coleman (1999) with a positive impact for the center leaders, which are also the more well-off
segment of the membership. The insignificant impact on poorer members was confirmed. On
the other hand, Hulme and Mosley (1996)using data from Indonesia, India, Bangladesh, and
Sri Lankafound a positive impact on income on average but like Coleman (2006) also found a
larger impact for the better-off members. Thus, the regressive result of this study is consistent
with some of the findings reported in the literature. This indicates that among poorer borrowers,
the availability of microcredit loans is not sufficient to ensure that the ultra poor invest in
sufficiently productive activities that can generate the income necessary to repay the loans and
earn them some profit.
It can be argued that the use of income quartile to determine the impact across socioeconomic
grouping done in the preceding analyses may be problematic because income is affected by the
treatment variable.21 To avoid the endogeneity problem, we re-estimated the above equations
using the education attainment of the reference person which is known to be directly related to
socioeconomic status and presumably acquired prior to the program. The summary of
estimation results are given in Table 21 (see Table A20 for the full estimation results). The
results substantially replicated the regressive impact. While the table shows that for those with
at most elementary education the impact on per capita income is negative, this is not statistically
significant. But the impacts for those with secondary and elementary education are significantly
positive and rising with higher education attainment.
21
We are grateful to D. Levine for pointing this out.
20
Table 21: Impact on Household Outcome by Education Status of Reference Person

Education status
At most elementary graduate
Secondary
At least some tertiary
Per capita income

Coeff.
Sig. Lev.
Coeff.
Sig. Lev.
-5,864
0.154
-1,511
0.619
9,727
0.009
7,302
0.008
9,805
0.034
4,206
0.218
Per capita sav 1

Coeff.
Sig. Lev.
-4,353
0.083
2,425
0.288
5,600
0.047
Per capita sav 2

Coeff.
Sig. Lev.
-4,904
0.060
3,521
0.138
6,353
0.030

Coeff.
Sig. Lev.
-45
0.962
1,567
0.074
2,254
0.037
Coeff. = Coefficients, Sig. Lev.. = Significance Level

IV.
SUMMARY, CONCLUSION AND RECOMMENDATIONS
The study used a quasi-experimental design (due to Coleman (1999)) to control for non-random
program participation and fixed-effects estimation to correct of non-random program placement.
In addition, it included former clients to correct for non-random attrition/drop-out problems which
were not considered in the original Coleman (1999) design. It also used recommended
estimation procedures for estimating average treatment effects described in Wooldridge (2002).
The survey results indicates that majority of the existing clients, the new clients, and the nonparticipating households which are deemed qualified for the program are not poor according to
the official definition. This is in sharp contrast to the other studies which indicated that majority
of the microfinance program clients are poor. Khandker (2003), for instance, indicates that 90%
of the microfinance program participants in Bangladesh in the 1991/92 survey and 70% in the
1998/99 survey are poor. Montgomery (2005) found that 70% of microfinance clients of the
Khushhali Bank in Pakistan are poor. Given these findings there is a need to re-examine the
targeting approach of the microfinance implementers in the Philippines. It could be that the
targeting approaches have the potential of identifying the desired clients but these are not just
implemented strictly. Another factor that needs to be looked at is whether there is enough
motivation for the implementers to seek poor clients. These have important implications on the
design of the microfinance programs.
The impact of the availability of program loans on per capita income is shown to be positive and
mildly significant. This is also true for per capita total expenditure and per capita food
expenditure. But it was also found that this impact is regressive, that is negative or insignificant
for poorer households and becomes only positive and increasing with richer households. This
negative or insignificant impact for poorer households and positive impact for richer households
provides some explanation of the mild significance of the coefficient for the total sample. It is
worth mentioning that this is not the only study that found a regressive impact. Coleman (2006)
using data from Thailand qualified the earlier no significant impact on consumption result in
Coleman (1999) with the finding of a positive impact for the center leaders which are also the
richer segment of the membership and that the insignificant impact is confined to poorer
members. Hulme and Mosley (1996) using data from Indonesia, India, Bangladesh and Sri
Lanka, on the other hand, found positive impact on income on average but in addition like
Coleman (2006) also found larger impact for better-off members. Thus the regressive result of
this study may not be entirely surprising but is certainly disturbing. This indicates that among
poorer borrowers the cost of and availability of program loans appears to be not sufficient to
prod them into selecting more productive activities that will not only pay the cost of borrowing
but also earn them some profit. One can also view this as the result of MFIs not screening
projects enough to have the desired results. This implies that attention to project selection must
also be an important component program design.
21
The program has enabled participants to reduce dependence on presumably higher priced nonGBA loans. In addition, it has increased the proportion of those having savings accounts in
program and other MFIs. It has also increased the amounts saved in those accounts. Together
these imply better consumption smoothing capabilities.
Another significant impact of the program is making program clients busier with larger number of
enterprises engaged in. This likewise resulted in bigger number of employed workers in these
enterprises. Given the thrust of the program to cater to micro-entrepreneurs, this result is hardly
surprising.
Finally, the study also found no significant impact on household assets as well as on human
capital investments such as health and education. It appears that the mild impact on income
and expenditures were not sufficient to drastically change either accumulation of household
assets or human capital investments.
In summary, the microfinance program has kept program clients economically active with more
enterprises and more employees. It has also improved consumption smoothing capabilities with
lesser dependence on presumably higher priced non-GBA loans and increased savings in both
program and non-program MFIs. Nonetheless, the impact on per capita income, total
expenditures and food expenditures is only mildly significant but with regressive features.
Considering the foregoing, for microfinance to be an effective poverty-alleviation tool there is a
need to review targeting procedures to know whether these are correctly identifying the
intended beneficiaries. There is also a need to regularly assess the economic status of clients to
avoid the drifting away from the focus on the poor and low income households. This cannot be
overemphasized considering that MFIs may not have sufficient motivation to seek out poor
clients. Finally, considering the regressive impact on income, there is a need to assist the poor
in improving the selection of projects so that these do not only ensure the payment of the loans
but also generate ample profit as well. Again in project selection the concern of the MFIs may
be limited to just ensuring repayment and not generating profits for their clients.
References
Amin S., A. S. Rai and G. Ropa (2003) Does microcredit reach the poor and vulnerable?
Evidence from Northern Bangladesh, Journal of Development Economics, 70(1), 59-82.
Armendariz de Aghion, B. and J. Morduch (2005) The Economics of Microfinance. Cambridge,
MA USA: MIT Press.
Bebczuk, R. and F. Haimovich (2007) MDGs and Microcredit: An Empirical Evaluation for Latin
American Countries, Working Paper No. 48. Universidad Nacional de la Plata, Argentina.
Coleman, B. (1999) The Impact of Group Lending in Northeast Thailand, Journal of
Development Economics, 60, 105-141.
Coleman, B. (2006) Microfinance in Northeast Thailand: Who Benefits and How Much? World
Development, 34(9) 1612-1638.
22
Karlan, D. (2001) Microfinance Impact Assessments: The Perils of Using New Members as
Control Group, Journal of Microfinance, December.
Khandker, S. (1998) Fighting Poverty with Microcredit: Experience from Bangladesh. New York:
Oxford University Press.
Khandker, S. (2003) Micro-Finance and Poverty: Evidence Using Panel Data from
Bangladesh, WB Policy Research Working Paper 2945.
Maldonado, J. (2005) The Influence of Microfinance on the Education Decisions of Rural
Households: Evidence From Bolivia, Bogota, Columbia: Universidad de los Andes CEDE
Document No 2005-46.
Meyer, R. (2002), Track Record of Financial Institutions Assisting the Poor in Asia ADB
Institute Research Paper No.49, Asian Development Bank.
Montgomery, H. (2005) Serving the Poorest of the Poor: The Poverty Impact of the Khushhali
Banks Microfinance Lending in Pakistan, ADB Institute.
Montgomery, H. and J. Weiss (2005) Great Expectations: Microfinance and Poverty Reduction
in Asia and Latin America, ADBI Research Paper Series No. 63.
Morduch, J. (1999) Does Microfinance Really Help the Poor? New Evidence from Flagship
Program in Bangladesh, Cambridge, MA: HIID and Hoover Institution.
Orr, L. (1997) Social Experimentation: Evaluating Public Programs with Experimental Methods.
Washington, DC: Department of Human Services, USA.
Papke, L. and J. Wooldridge (1996) Econometric Methods for Fractional Response Variables
with an Application to 401(k) Plan Participation Rates, Journal of Applied Econometrics 11,
619-632.
Pitt, M. and S. Khandker (1998) The Impact of Group-Based Credit Programs on Poor
Households in Bangladesh: Does the Gender of Participants Matter? Journal of Political
Economy 106: 958-96.
Pitt, M. (1999) Reply to Jonathan Morduchs Does Microfinance Really Help the Poor?
New Evidence from Flagship Programs in Bangladesh Brown University, available at:
http://www.pstc.brown.edu/~mp/reply.pdf.
Pitt, M., S. Khandker, O. Chowdhury and D. Millimet. (2003) Credit Programs for the Poor and
the Health Status of Children in Rural Bangladesh International Economic Review, 44 (1) 87118.
Pitt, M.M. and Khandker, S. (1998) The Impact of Group-based Credit Programs on Poor
Households in Bangladesh: Does the Gender of Participants Matter? Journal of Political
Economy, 106(2), 958-996.
Rosenbaum, P. and D. Rubin (1983) The Central Role of Propensity Score in Observational
Studies for Causal Effects, Biometrika 70, 41-55.
23
Weiss, J. H. Montgomery and E. Kurmanalieva (2003) Micro Finance and Poverty Reduction in
Asia: What is the Evidence? ADBI Research Paper Series No. 53.
Wooldridge, J (2002) Econometric Analysis of Cross Section and Panel Data. Cambridge, MA,
USA: MIT Press.
Zeller, M., M. Sharma, A. Ahmed and S. Rashid (2001) Group-Based Financial Institutions for
the Rural Poor in Bangladesh: An Institutional- and Household-Level Analysis. International
Food Policy Research Institute (IFPRI).
24
Annex Estimation Results
Table A1: Impact on per capita income
Estimation: Fixed-effects regression
Variables
Member
Availed loan
Age, reference person (rp)
Age square, rp
Female, rp
Elementary, rp
Secondary, rp
Tertiary, rp
Years in village
House size
Availed loan*(demeaned\a age)
Availed loan*(demeaned\a age square)
Availed loan*(demeaned\a female rp)
Availed loan*(demeaned\a elementary rp)
Availed loan*(demeaned\a secondary rp)
Availed loan*(demeaned\a tertiary rp)
Availed loan*(demeaned\a years in vill.)
Availed loan*(demeaned\a house size)
Months program available
Mos prog avail*(demeaned\a age)
Mos prog avail*(demeaned\a age square)
Mos prog avail*(demeaned\a female rp)
Mos prog avail*(demeaned\a elementary rp)
Mos prog avail*(demeaned\a secondary rp)
Mos prog avail*(demeaned\a tertiary rp)
Mos prog avail*(demeaned\a years in village)
Mos prog avail*(demeaned\a house size)
Total loans availed (000)
Tot. loans av.*(demeaned\a age)
Tot. loans av.*(demeaned\a age square)
Tot. loans av.*(demeaned\a female rp)
Tot. loans av.*(demeaned\a elementary rp)
Tot. loans av.*(demeaned\a secondary rp)
Tot. loans av.*(demeaned\a tertiary rp)
Tot. loans av.*(demeaned\a years in village)
Tot. loans av.*(demeaned\a house size)
Number of loan cycles
No. loan cyls*(demeaned\a age)
No. loan cyls*(demeaned\a age square)
No. loan cyls*(demeaned\a female rp)
No. loan cyls*(demeaned\a elementary rp)
No. loan cyls*(demeaned\a secondary rp)
No. loan cyls*(demeaned\a tertiary rp)
No. loan cyls*(demeaned\a years in village)
No. loan cyls*(demeaned\a house size)
Constant
Model Statistics
Sample
F on Ho: u_i=0
Overall R-square
\a=variable-mean(variable)
Availed Loan
(1=Yes)
Coefficient t-value
-9.89
0.00
5221.98
1.65
-898.72
-1.92
13.93
2.72
15180.70
5.58
9488.92
0.8
15328.07
1.29
27422.61
2.29
57.27
0.76
78.60
4.39
757.41
0.61
-7.62
-0.58
3797.82
0.65
-27335.59
-1.11
-16851.55
-0.68
-28631.67
-1.15
-91.43
-0.62
-39.15
-1.46
Treatment variable
Months program
Total Loans
No. of loan
available
('000)
Cycles
Coefficient t-value Coefficient t-value Coefficient t-value
1087.43
0.51
749.49
0.33
804.41
0.35
-795.13
12.84
16775.15
9428.64
16412.28
27261.75
75.16
71.34
38.98
2.21
-0.01
-35.85
-310.79
-220.76
-319.36
-2.33
-0.32
-1.72
-820.93
2.55
12.69
6.32 15959.82
0.79 4567.34
1.38 9984.43
2.28 20929.58
1.02
-19.42
4.24
62.13
-1.80
2.55
6.09
0.39
0.85
1.77
-0.27
3.86
-799.90
12.79
16711.22
10509.13
16762.23
28313.50
21.11
75.55
-1.73
2.53
6.25
0.89
1.42
2.38
0.29
4.51
425.83
25.23
-0.03
-442.53
-4821.63
-3395.11
-4762.92
5.02
-4.00
28099.96
0.85
0.17
-0.02
-0.61
-1.27
-0.90
-1.25
0.27
-1.58
1.83
1.01
0.14
-0.06
-0.54
-1.07
-0.76
-1.09
-1.30
-1.02
20.25
11.48
-0.07
11.06
-159.88
38.64
-38.99
3.52
-0.04
30650.25
1.97 27756.76
1.79 37483.58
2018
3.649
0.099
2013
3.62
0.098
2013
3.54
0.114
0.38
0.77
-0.46
0.16
-0.32
0.08
-0.08
2.01
-0.22
2.45
2013
3.586
0.103
25
Table A2: Impact on per capita expenditure
Variables
Member
Availed loan
Age square, rp
Female, rp
Elementary, rp
Secondary, rp
Tertiary, rp
Years in village
House size
Constant
Model Statistics
Sample
F on Ho: u_i=0
Overall R-square
Treatment variable
Availed Loan
Months program
Total Loans
No. of loan
(1=Yes)
available
('000)
Cycles
Coefficient
t-value Coefficient t-value Coefficient t-value Coefficient t-value
-2245.97
-1.37 -1392.94
-0.89 -1656.46
-0.99 -1604.32
-0.95
4135.51
1.77
-599.17
-1.73
-571.02
-1.67
-547.01
-1.62
-587.04
-1.72
8.75
2.31
8.65
2.32
8.32
2.26
8.82
2.36
8709.82
4.34 9790.94
4.99
8781.84
4.51 9552.95
4.83
5809.69
0.66 6000.83
0.68
2723.40
0.31 6304.99
0.72
9608.48
1.09 10542.87
1.20
7145.72
0.82 10700.61
1.23
19907.30
2.25 19893.05
2.25 15989.76
1.82 19962.39
2.27
49.74
0.89
55.49
1.02
-4.20
-0.08
24.74
0.46
79.15
5.99
73.49
5.91
61.85
5.18
76.51
6.18
453.59
0.50
-4.04
-0.42
5071.44
1.18
-14956.79
-0.82
-9469.61
-0.52
-22679.89
-1.23
-73.88
-0.68
-57.10
-2.88
31.34
1.10
3.71
0.31
-0.04
-0.29
12.24
0.25
-170.34
-0.79
-132.93
-0.62
-261.81
-1.20
-1.47
-1.11
-0.61
-2.62
28.79
0.72
7.52
0.68
-0.07
-0.58
82.87
1.59
-122.18
-0.33
-75.32
-0.20
-146.74
-0.39
2.36
1.81
-0.14
-1.00
371.66
1.01
52.92
0.48
-0.51
-0.44
203.62
0.38
-2604.32
-0.93
-2104.61
-0.75
-3335.44
-1.18
2.05
0.15
-5.76
-3.08
24014.61
2.09 22545.69
1.97 27587.03
2.43 23135.17
2.03
2018
3.776
0.097
2013
3.759
0.095
2013
3.659
0.103
2013
3.728
0.097
26
Table A3: Impact on percapita savings (Income-Total Expenditure)
Treatment variable
Availed Loan
Months program
Total Loans
No. of loan
(1=Yes)
available
('000)
Cycles
Coefficient t-value Coefficient t-value Coefficient t-value Coefficient t-value
Member
2236.09
1.64 2480.37
1.91 2405.95
1.74 2408.72
1.71
Availed loan
1086.47
0.56
-299.55
-1.04
-224.11
-0.79
-273.93
-0.98
-212.86
-0.75
Age square, rp
5.18
1.64
4.19
1.35
4.37
1.43
3.96
1.27
Female, rp
6470.88
3.86 6984.20
4.28 7177.98
4.44 7158.27
4.35
Elementary, rp
3679.23
0.50 3427.81
0.47 1843.94
0.25 4204.14
0.58
Secondary, rp
5719.60
0.78 5869.42
0.80 2838.71
0.39 6061.62
0.83
Tertiary, rp
7515.31
1.02 7368.70
1.00 4939.82
0.68 8351.10
1.14
Years in village
7.53
0.16
19.67
0.43
-15.21
-0.35
-3.63
-0.08
House size
-0.55
-0.05
-2.15
-0.21
0.28
0.03
-0.96
-0.09
303.82
0.40
-3.57
-0.44
-1273.61
-0.36
Availed loan*(demeaned\a elementary rp) -12378.80
-0.81
-7381.94
-0.49
-5951.78
-0.39
-17.56
-0.19
17.94
1.09
7.64
0.32
-1.50
-0.15
0.03
0.25
-48.09
-1.17
-140.45
-0.78
-87.83
-0.49
-57.55
-0.32
-0.86
-0.78
0.29
1.49
-8.54
-0.26
3.97
0.43
0.00
-0.04
-71.81
-1.65
-37.70
-0.12
113.95
0.37
107.76
0.35
1.16
1.07
0.10
0.85
54.17
0.18
-27.70
-0.30
0.48
0.49
-646.14
-1.45
-2217.31
-0.95
-1290.50
-0.55
-1427.48
-0.61
2.97
0.26
1.77
1.13
Constant
6635.64
0.69 5211.08
0.55 9896.54
1.05 4964.79
0.52
Model Statistics
Sample
2018
2013
2013
2013
F on Ho: u_i=0
2.027
2.005
2.024
2.002
Overall R-square
0.033
0.035
0.043
0.036
Variables
27
Table A4: Impact on percapita savings (Income-Total Exp+Educ+Health+Dur. Fur.)
Treatment variable
Availed Loan
Months program
Total Loans
No. of loan
(1=Yes)
available
('000)
Cycles
Member
1666.10
1.17 1925.45
1.42 1736.91
1.21
1761.29
1.20
Availed loan
1626.08
0.80
-221.99
-0.74
-125.61
-0.42
-166.51
-0.57
-116.60
-0.39
Age square, rp
5.16
1.57
3.94
1.22
4.00
1.26
3.72
1.15
Female, rp
6461.64
3.71 7189.74
4.24 6994.08
4.17
7190.84
4.21
Elementary, rp
3950.23
0.52 3615.23
0.48 1617.78
0.21
4512.22
0.60
Secondary, rp
6514.62
0.86 6749.56
0.89 3187.02
0.42
6999.33
0.93
Tertiary, rp
9577.25
1.25 9451.53
1.24 6525.52
0.86
10497.16
1.38
Years in village
16.86
0.35
30.55
0.65
-14.76
-0.32
3.94
0.08
House size
7.19
0.63
4.15
0.39
7.42
0.72
7.83
0.73
480.18
0.61
-5.70
-0.68
41.22
0.01
-12821.71
-0.81
-6660.57
-0.42
-6814.94
-0.43
-30.88
-0.33
19.74
1.15
14.23
0.58
-0.79
-0.08
0.02
0.14
-43.30
-1.01
-142.93
-0.77
-81.83
-0.44
-70.21
-0.37
-1.17
-1.02
0.35
1.76
0.72
0.02
3.87
0.41
-0.01
-0.05
-40.97
-0.91
-17.05
-0.05
150.83
0.47
126.77
0.39
1.51
1.34
0.12
1.02
154.39
0.48
-24.89
-0.26
0.43
0.42
-503.39
-1.09
-2288.08
-0.94
-1238.19
-0.51
-1571.55
-0.64
1.63
0.14
1.64
1.01
Constant
4909.24
0.49 3030.73
0.31 8276.83
0.85
2743.90
0.28
Model Statistics
Sample
2018
2013
2013
2013
F on Ho: u_i=0
2.123
2.095
2.114
2.09
Overall R-square
0.048
0.049
0.061
0.052
Variables
28
Table A5: Impact on percapita food expenditure
Treatment variable
Availed Loan
Months program
Total Loans
No. of loan
(1=Yes)
available
('000)
Cycles
Member
-618.85
-1.19
-382.19
-0.77
-434.77
-0.82
-446.90
-0.83
Availed loan
1332.99
1.80
-342.24
-3.12
-328.97
-3.04
-316.48
-2.96
-334.18
-3.09
Age square, rp
4.39
3.66
4.31
3.65
4.15
3.55
4.36
3.69
Female, rp
2565.08
4.03 2710.60
4.37 2635.84
4.27 2748.78
4.39
Elementary, rp
199.15
0.07
229.19
0.08
-335.95
-0.12
419.52
0.15
Secondary, rp
1773.31
0.64 1962.27
0.71 1504.31
0.54 2153.23
0.78
Tertiary, rp
4461.22
1.59 4422.55
1.58 3739.61
1.34 4428.32
1.59
Years in village
-16.19
-0.92
-12.78
-0.74
-17.38
-1.03
-17.40
-1.01
House size
8.84
2.11
8.07
2.05
5.20
1.37
7.74
1.97
446.99
1.55
-4.78
-1.56
1843.03
1.36
-3213.50
-0.56
-3022.67
-0.52
-4966.67
-0.85
6.22
0.18
-13.10
-2.09
12.09
1.35
4.63
1.24
-0.05
-1.36
19.13
1.22
-35.83
-0.53
-42.77
-0.63
-55.73
-0.81
-0.14
-0.33
-0.16
-2.13
13.55
1.07
4.46
1.27
-0.05
-1.37
26.11
1.58
-43.94
-0.37
-59.57
-0.51
-58.30
-0.49
0.29
0.70
-0.04
-0.91
140.92
1.20
50.70
1.45
-0.58
-1.56
161.53
0.95
-674.11
-0.76
-747.03
-0.84
-767.18
-0.86
2.00
0.47
-1.17
-1.98
Constant
16543.55
4.55 16011.16
4.42 16588.78
4.60 16076.59
4.46
Model Statistics
Sample
2018
2013
2013
2013
F on Ho: u_i=0
3.597
3.572
3.532
3.576
Overall R-square
0.061
0.06
0.063
0.059
Variables
29
Table A6: Impact on other (non-GBA) loans
Availed of NonGBA Loans
Coefficient z-value
Variables
Member
Availed loan
Propensity score (PS)
Availed loan.*(demeaned\a PS)
Sigma
Constant
Sample
Chi-square
Estimation procedure
-0.031
-0.175
1.160
-1.170
-0.42
-1.91
2.68
-1.25
-0.903
-8.40
2001
13.823
Probit
Amt. of NonGBA Loans

Coefficient z-value
-3.926
-0.654
46.546
-60.886
35.335
-35.813
2018
Tobit
No. of NonGBA Loans

Coefficient z-value
-1.57
-0.21
3.03
-1.98
28.07
-8.80
0.022
0.166
0.348
-0.819
478
2.329
Poisson
Table A7: Probit first stage - propensity score
Variables
Coefficient
Standard
Error
0.0770
-0.0007
-0.1447
0.0643
0.0543
0.0434
0.0035
0.0020
-2.9919
0.0190
0.0002
0.0908
0.3974
0.3962
0.3993
0.0023
0.0005
0.5924

Age square., rp
Female, rp
Elementary, rp
Secondary, rp
Tertiary, rp
Years in village
House size
Constant
Sample
Pseudo R-square
z-value
4.06
-3.33
-1.59
0.16
0.14
0.11
1.53
4.35
-5.05
2018
0.034
Table A8: Impact on personal savings
Variables
Member
Availed loan
Cut point 1
Cut point 2
Constant
Sample
Chi-square
Have personal
Amount of personal
savings
saving by group
Coefficient z-value Coefficient z-value
0.341
4.95
-0.447
-4.7
0.687
7.20
0.347
3.09
0.465
1.13
0.990
1.83
-1.469
-1.55
-0.710
-0.67
0.630
4.56
1.143
8.05
-0.045
-0.44
2010
154.81
Probit
1056
25.37
Ordered Probit
0.18
0.96
0.48
-0.68
30
Table A9: Impact on household enterprise and employment
Estimation: Fixed-Effects Poisson Regression
Total Number of
Total number of
enterprises
employees
Coefficient z-value
IRR
Coefficient z-value
IRR
Member
0.210
4.13
1.23
0.237
5.51
1.27
Availed loan
0.185
2.63
1.20
0.161
2.73
1.17
0.850
2.95
2.34
0.814
3.36
2.26
-0.273
-0.54
0.76
0.294
0.74
1.34
Variables
Sample
Chi-square
IRR = Incidence Rate Ratio
2018
91.544
2018
156.787
Table A10: First Stage Propensity Score Regression

Dependent Variable: Avail loan
Estimation Procedure: Probit
Independent Variables
Age square, rp
Female, rp
Elementary, rp
Secondary, rp
Tertiary, rp
Years in village
House size
Constant
Sample Size
LR Chi-square (8)
Coefficient
Standard
Error
0.077
-0.001
-0.145
0.064
0.054
0.043
0.004
0.002
-2.992
0.0190
0.0002
0.0908
0.3974
0.3962
0.3993
0.0023
0.0005
0.5924
2,018
74.73
z-value
4.06
-3.33
-1.59
0.16
0.14
0.11
1.53
4.35
-5.05
31
Table A11: Impact on Total Assets
Estimation: Fixed-effects Tobit
Variables
Coef.
Std. Err.
Member
Availed loan
Constant
1015.2 100937.5
-37891.2 122070.3
1396583.0 612817.9
-975040.5 1208059.0
-138725.2 150595.1
sigma_u
sigma_e
0.0
1866553.0
rho
45864.64
0.0
Sample
Chi-square (4)
2018
5.38
Table A12: First Stage Propensity Score Regression

Dependent Variable: Avail loan
Estimation Procedure: Probit
Coef.
Age square, rp
Female, rp
Elementary, rp
Secondary, rp
Tertiary, rp
Years in village
House size
Constant
Sample Size
LR Chi-square (8)
0.0770
-0.0007
-0.1447
0.0643
0.0543
0.0434
0.0035
0.0020
-2.9919
2018
74.73
Standard
Error
0.0190
0.0002
0.0908
0.3974
0.3962
0.3993
0.0023
0.0005
0.5924
z-value
4.06
-3.33
-1.59
0.16
0.14
0.11
1.53
4.35
-5.05
z-value
0.01
-0.31
2.28
-0.81
-0.92
0.00
32
Table A13: Impact on Education
Age groups
Exp. per
6-12
13-16
17-24
att. child
Coefficient z-value Coefficient z-value Coefficient z-value Coefficient t-value
Member
-0.044
-0.16
-0.049
-0.21
0.017
0.1
-39.305
-0.05
Availed loan
0.572
1.42
0.144
0.52
0.24
1.22 -681.985
-0.62
2.135
1.1
3.035
1.52
5.807
6.01
0.209
0.06
1.163
0.4
-4.817
-2.26
104.755
0.53
Age square, rp
1.499
0.67
Female, rp
-575.653
-0.51
Elementary, rp
1795.53
0.96
Secondary, rp
5082.852
2.63
Tertiary, rp
6394.019
3.03
Years in village
27.014
0.92
House size
16.096
1.98
Elem school available
853.967
0.61
Secondary school available
-2476.33
-1.61
Tertiary school available
1429.158
0.78
725.98
1.8
-8.19
-1.94
-505.695
-0.26
-1859.15
-0.46
-1175.27
-0.28
-3918.33
-0.91
Availed loan*(demeaned\a years in village)
-107.584
-2.2
-9.484
-0.97
Availed loan*(demeaned\a elementary school)
-223.88
-0.09
Availed loan*(demeaned\a secondary school)
2162.682
1.22
Availed loan*(demeaned\a tertiary school)
132.582
0.04
Constant
2.392
4.88
1.139
2.3
-2.245
-8.9 -6009.35
-1.21
Variables
Sample
1036
Chi-square
5.261
R-square
GLM\b
\b=Fractional Logit (Papke and Wooldridge,1996)
758
7.125
GLM\b
868
39.436
GLM\b
Table A14: First stage probit - Education
Variable
Age square, rp
Female, rp
Elementary, rp
Secondary, rp
Tertiary, rp
Years in village
House size
Elem school available
Secondary school available
Tertiary school available
Constant
Sample
Pseudo R-square
Coefficient
Standard
Error
0.0781
-0.0007
-0.1493
0.0671
0.0537
0.0382
0.0033
0.0019
0.0590
-0.0432
0.1923
-3.0729
0.0191
0.0002
0.0912
0.3974
0.3963
0.3994
0.0023
0.0005
0.1207
0.0716
0.1057
0.6073
2001
0.0345
z-value
4.09
-3.33
-1.64
0.17
0.14
0.10
1.41
4.11
0.49
-0.60
1.82
-5.06
1404
0.088
Fixed-Eff.
33
Table A15: Impact on health
Proportion
ill or injured
Variables
Member
Availed loan
Age square, rp
Female, rp
Elementary, rp
Secondary, rp
Tertiary, rp
Years in village
House size
Govt hospital available
Private hospital available
Private clinic available
Health clinic available
Barangay Health Station available
Availed loan*(demeaned\a age, rp)
Availed loan*(demeaned\a age square, rp)
Availed loan*(demeaned\a years in village)
Availed loan*(demeaned\a govt hospital)
Availed loan*(demeaned\a private hospital)
Availed loan*(demeaned\a private clinic)
Availed loan*(demeaned\a health clinic)
Availed loan*(demeaned\a BHS)
Constant
Coefficient z-value
0.193
1.39
0.062
0.38
1.967
2.17
-1.828
-1.15
-2.927
-13.16
Sample
1994
Chi-square
9.199
R-square
GLM\b
\b=Fractional Logit (Papke and Wooldridge,1996)
Prop who seek

Prop fully
Per capita
treatment if
immunized
medical
ill or injured
0-5 years
expenditures
Coefficient z-value Coefficient z-value Coefficient t-value
0.171
0.73
0.259
1.34 -253.469
-1.56
-0.035
-0.13
-0.031
-0.12
19.656
0.07
2.17
1.62
0.967
0.92
0.256
0.12
-1.921
-0.9
-174.387
-1.98
2.3
2.11
5.883
0.02
485.616
1.05
202.958
0.55
493.215
1.33
4.781
0.69
2.226
1.01
-252.641
-0.37
479.518
0.81
-398.245
-1.11
333.16
1.67
83.307
0.68
63.685
0.53
-0.943
-0.65
230.852
0.48
-736.62
-0.98
-178.779
-0.28
-997.452
-1.36
-10.416
-0.75
-1.76
-0.7
-718.734
-0.68
-155.636
-0.2
335.812
0.69
(dropped)
-320.088
-0.95
0.196
0.56
0.464
1.96 2941.841
1.74
456
5.448
GLM\b
794
3.733
GLM\b
Table A16: First stage probit - Health
Variable
Age square, rp
Female, rp
Elementary, rp
Secondary, rp
Tertiary, rp
Years in village
House size
Govt hospital available
Private hospital available
Private clinic available
Barangay Health Station available
Constant
Sample
Pseudo R-square
Coeficient
Standard
Error
0.0801
-0.0007
-0.1466
0.0954
0.0950
0.0765
0.0036
0.0019
0.4778
-0.1462
-0.1156
-1.4195
-3.1444
0.0193
0.0002
0.0920
0.3956
0.3946
0.3976
0.0023
0.0005
0.1144
0.1507
0.0900
0.0860
0.6023
1994
0.042
z-value
4.15
-3.43
-1.59
0.24
0.24
0.19
1.55
4.10
4.18
-0.97
-1.28
-16.51
-5.22
1994
0.031
Fixed-Eff.
34
Table A17: Impact on hunger and reduced food consumption
Variable
Member
Availed loan
Constant
Sample
Chi-square
Hunger
Reduced food
incidence
incidence
Coefficient z-value Coefficient z-value
-0.049
-0.32
-0.023
-0.26
-0.004
-0.02
-0.051
-0.47
1.292
1.55
-0.343
-0.63
-0.041
-0.03
1.607
1.66
-2.287
-10.96
-1.114
-8.49
2014
3.645
Probit
2009
3.042
Probit
Table A18: First stage probit - Hunger
Variable
Age square, rp
Female, rp
Elementary, rp
Secondary, rp
Tertiary, rp
Years in village
House size
Constant
Sample
Pseudo R-square
Coefficient
Standard
Error
0.0770
-0.0007
-0.1447
0.0643
0.0543
0.0434
0.0035
0.0020
-2.9919
0.0190
0.0002
0.0908
0.3974
0.3962
0.3993
0.0023
0.0005
0.5924
2,018
0.0340
z-value
4.06
-3.33
-1.59
0.16
0.14
0.11
1.53
4.35
-5.05
35
Table A.19: Impact on per capita income, expenditure, savings, food expenditure with income quintile
Per capita income
Coeff.
t-value
Variables
Member
Quartile 1*Availed Loan
Age, ref. person (rp)
Age square, rp
Female, rp
Elem, rp
Secondary, rp
Tertiary, rp
Years in village
House size
Constant
Model Statistics
Sample
F on Ho: u_i=0
Overall R2
-116.63
-23,213.95
-13,903.13
-1,212.17
45,113.72
-882.79
13.71
14,827.07
6,826.59
13,450.09
25,738.28
57.94
80.01
1,434.61
-19.40
-3,305.85
-14,132.79
-15,374.72
-30,681.25
37.58
-71.70
32,381.63
-0.06
-5.08
-3.34
-0.30
11.35
-2.00
2.84
5.78
0.61
1.20
2.28
0.82
4.74
1.23
-1.57
-0.60
-0.61
-0.66
-1.31
0.27
-2.83
2.21

Coeff.
t-value
Per capita sav 1

Coeff.
t-value
-2,293.17
-9,459.75
-6,752.58
1,849.64
23,915.60
-592.84
8.65
8,531.91
4,464.90
8,664.14
19,060.37
49.95
79.93
782.92
-9.85
1,620.92
-8,641.05
-9,062.08
-24,004.88
-6.21
-73.28
24,922.72
-1.44
2,176.54
-2.71 -13,754.20
-2.13 -7,150.55
0.60 -3,061.81
7.88 21,198.11
-1.76
-289.95
2.35
5.05
4.36
6,295.16
0.52
2,361.69
1.01
4,785.95
2.21
6,677.91
0.92
7.99
6.20
0.08
0.88
651.69
-1.04
-9.55
0.38 -4,926.78
-0.49 -5,491.74
-0.51 -6,312.65
-1.34 -6,676.38
-0.06
43.78
-3.78
1.58
2.23
7,458.91
2,018
3.437
0.157
2,018
1.652
0.122
2018
3.068
0.223
Per capita sav 2

Coeff.
t-value
1.65
-4.79
-2.74
-1.21
8.48
-1.04
1.66
3.90
0.33
0.68
0.94
0.18
0.01
0.89
-1.23
-1.42
-0.37
-0.43
-0.45
0.50
0.10
0.81
1,600.41
-14,567.24
-7,680.61
-3,010.56
23,928.61
-211.35
5.01
6,265.26
2,489.98
5,478.81
8,648.25
17.34
7.89
863.49
-12.30
-3,952.96
-5,197.95
-5,467.91
-7,609.41
37.61
1.60
5,824.55

Coeff.
t-value
1.18
-4.91
-2.84
-1.15
9.27
-0.74
1.60
3.76
0.34
0.75
1.18
0.38
0.72
1.14
-1.53
-1.10
-0.34
-0.36
-0.50
0.42
0.10
0.61
2,018
1.723
0.147
-630.41
-3,476.87
-1,408.46
1,382.08
6,659.93
-341.30
4.38
2,523.97
-180.25
1,514.57
4,229.01
-16.06
9.05
545.47
-6.47
646.84
-1,560.10
-3,118.41
-5,532.37
23.85
-17.51
16,801.31
-1.24
-3.13
-1.39
1.41
6.89
-3.18
3.73
4.05
-0.07
0.56
1.54
-0.93
2.20
1.92
-2.15
0.48
-0.28
-0.55
-0.97
0.71
-2.84
4.72
2,018
3.232
0.114
\a=var-mean(var)
Table A.20: Impact on per capita income, expenditure, savings, food expenditure by education of reference person
Variable
Member
At most some elementary
Secondary
At least some tertiary
Age square, rp
Female, rp
Years in village
House size
Availed loan*(demeaned\a age sq
Availed loan*(demeaned\a female
Availed loan*(demeaned\a years i
Availed loan*(demeaned\a house
Constant
Model Statistics
Sample
F on Ho: u_i=0
Overall R2
Per capita income

Coeff.
t-value
41.82
-5,864.09
9,726.60
9,805.45
-861.01
12.37
15,908.65
18.61
91.32
738.29
-6.39
3,184.27
-58.74
-51.59
48,203.67
2,018
3.966
0.069
0.02
-1.43
2.60
2.12
-1.82
2.40
5.81
0.25
5.10
0.59
-0.48
0.54
-0.40
-1.91
4.61

Coeff.
t-value
-2,222.14
-1,511.02
7,301.51
4,205.85
-577.27
7.64
9,348.97
19.88
89.62
441.37
-3.13
4,536.02
-50.10
-67.07
35,872.44
2,018
4.113
0.062
-1.34
-0.50
2.64
1.23
-1.65
2.00
4.62
0.35
6.76
0.48
-0.32
1.05
-0.46
-3.36
4.64
Per capita sav 1

Coeff.
t-value
2,263.96
-4,353.08
2,425.09
5,599.61
-283.74
4.73
6,559.68
-1.26
1.71
296.92
-3.26
-1,351.75
-8.64
15.48
12,331.24
2,018
2.086
0.028
1.66
-1.74
1.06
1.99
-0.98
1.50
3.93
-0.03
0.16
0.39
-0.40
-0.38
-0.10
0.94
1.93
Per capita sav 2

Coeff.
t-value
1,700.56
-4,903.94
3,520.61
6,353.41
-203.46
4.57
6,634.47
4.27
10.78
470.60
-5.26
-98.17
-19.17
16.03
11,669.44
2,018
2.223
0.039
1.20
-1.88
1.48
2.17
-0.68
1.39
3.82
0.09
0.95
0.59
-0.63
-0.03
-0.20
0.94
1.76

Coeff.
t-value
-589.73
-45.32
1,566.66
2,254.07
-333.74
4.02
2,751.31
-25.45
11.92
442.62
-4.48
1,693.89
13.81
-16.10
18,775.29
2,018
3.875
0.03
-1.13
-0.05
1.79
2.09
-3.02
3.33
4.30
-1.43
2.85
1.52
-1.45
1.24
0.40
-2.55
7.68

Pidsdps0805 MF Rural Households

Diunggah oleh

Informasi Dokumen

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Pidsdps0805 MF Rural Households

Diunggah oleh

Hak Cipta:

Format Tersedia

Philippine Institute for Development Studies

Surian sa mga Pag-aaral Pangkaunlaran ng Pilipinas

The PIDS Discussion Paper Series

IMPACT OF MICROFINANCE ON RURAL HOUSEHOLDS

IMPACT OF MICROFINANCE ON RURAL HOUSEHOLDS

FRAMEWORK, SURVEY DESIGN AND DATA, ESTIMATION PROCEDURE

From Table 1, impact is given by the expression:

Yij = F ( 1 X ij + 2V j + 3Mij + 4Tij + ij )

This has also been rendered as

Total No. of Borrowers

Respondent Characteristics and Outreach

Source: Operations Evaluation Mission

Age of reference person (rp)

***, **, * - significant at 1%, 5% and 10%, respectively; two-tailed hypothesis

Table 6: Basic Welfare Indicators of Respondents

***, **, * - significant at 1%, 5% and 10%, respectively; two-tailed hypothesis

Source: Operations Evaluation Mission.

Source: Operations Evaluation Mission

Figure 3: Distribution of Sample Qualified Nonparticipating Households

Source: Operations Evaluation Mission.

Source: Operations Evaluation Mission

Estimation Results - Impact of Microfinance

Impact on per capita income, expenditure, savings, and expenditure on food

Availed non-GBA loans

Source: Operations Evaluation Mission

Table 9: Impact on non-GBA loans

Availed non-GBA loans

Source: Operations Evaluation Mission

***, **, * - significant at 1%, 5% and 10%, respectively; two-tailed hypothesis

Source: Operations Evaluation Mission

Impact on the Number of Enterprises and Employment

With household enterprise

Source: Operations Evaluation Mission

Total number of enterprises

Source: Operations Evaluation Mission

With agriculture and comm land

Source: Operations Evaluation Mission

Impact on human capital investments (education and health)

With children 6-12 years old

***, **, * - significant at 1%, 5% and 10%, respectively; two-tailed hypothesis

Table 16: Health outcomes

Proportion of members ill/injured

Impact by Different Socioeconomic Groups

Per Capita Income

Coeff. = Coefficients, Sig. Lev.. = Significance Level

Source: Operations Evaluation Mission

Source: Operations Evaluation Mission

We are grateful to D. Levine for pointing this out.

Table 21: Impact on Household Outcome by Education Status of Reference Person

Per capita income

Per capita sav 1

Per capita sav 2

Per capita food exp.

Coeff. = Coefficients, Sig. Lev.. = Significance Level

SUMMARY, CONCLUSION AND RECOMMENDATIONS

Amt. of NonGBA Loans

No. of NonGBA Loans

Table A7: Probit first stage - propensity score

Age, reference person (rp)

Table A8: Impact on personal savings

Table A10: First Stage Propensity Score Regression

Table A12: First Stage Propensity Score Regression

Table A14: First stage probit - Education

Prop who seek

, , - significant at 1%, 5% and 10%, respectively; two-tailed hypothesis

, , - significant at 1%, 5% and 10%, respectively; two-tailed hypothesis

, , - significant at 1%, 5% and 10%, respectively; two-tailed hypothesis

, , - significant at 1%, 5% and 10%, respectively; two-tailed hypothesis